Engineering, product, and business share a building and read different calendars. Each area decides from the number in front of it, and the missing number sits on someone else's desk. The quarter closes on code for changed requirements, a feature nobody uses, and two months of team time on an integration nobody asked for.
Each area holds a number the others overlook
The engineer pushes to refactor the payment system: technical debt drops uptime in six months. The product manager pushes the onboarding feature: retention has fallen for two quarters. The CEO pushes the custom integration: two large contracts depend on it. All three are right, and none of them swapped documents.
The engineer does not see that onboarding closes a R$ 500k contract, and the product manager does not see that the refactor prevents a 40% drop in uptime. The CEO gets the two-month engineering estimate after the contract is signed.
Rituals that put the three in one room
- A quarterly cross-roadmap review. Engineering, product, and business open the three roadmaps on one screen. Each area shows its priority, the trade it accepts, and the dependency it cannot resolve alone.
- A shared impact matrix. Every initiative carries technical impact, product impact, and business impact on one line. Priority turns into a line comparison.
- A weekly 30-minute check-in. One representative per area and a fixed agenda: new blocker, contested priority, pending decision. A blocker found on Monday is gone before it becomes rework on Friday.
- A decision record. Context, discarded options, and rationale go in writing. Michael Nygard calls the document an Architecture Decision Record in "Documenting Architecture Decisions" (2011): the forces at play and the single decision that answers them.
Translate what the other side hears
Latency, throughput, and availability say nothing to someone watching churn. Write the sentence twice: "we take the API from 200ms to 50ms" and "checkout stops freezing and churn drops 5%". Budget comes out of the second one.
Product joins the conversation the day the business need appears, before the specification. An engineer invited late returns a cost against a closed requirement; an engineer invited early returns an option that costs less.
Escalation without blame
When the three areas stall at the operational level, the conflict moves up to whoever decides, with the written matrix attached. Use the path to settle it in days. Without it, the team that raised the blocker absorbs the rework.
Marty Cagan argues in Inspired (Wiley, 2017) that product and engineering share one queue and decide together. With separate queues, each area optimizes its own number and the quarter closes with rework.
You align the three areas by publishing one matrix for all of them. What survives the meeting is what you wrote before it.
